Net Trap EA - variant backtesting

This is a discussion on Net Trap EA - variant backtesting within the Forex Net Trap forums, part of the T2W Archive category; Originally Posted by runningdeer Hello Spin, Thanks for making all your research available. Especially to novices like myself. Can you ...

Reply
 
LinkBack Thread Tools Search this Thread
Old Feb 2, 2011, 4:24pm   #31
Joined Dec 2006
Re: Net Trap EA - variant backtesting

SpindriftUK started this thread
Quote:
Originally Posted by runningdeer View Post
Hello Spin,

Thanks for making all your research available. Especially to novices like myself. Can you help me understand a couple of points on your spreadsheets. The figures under maximal drawdowns I assume relate to the amount required to cover the largest number of consecutive losses. Is that right? And what does the absolute drawdown figure relate to?

Many thanks, RD
They're both numbers calculated by the Strategy Tester. Below are the definitions from the MT4 manual. (The definition for Maximal Drawdown loses a bit in the translation from the Russian I think )
Absolute drawdown — the largest loss is below initial deposit value;
Maximal drawdown — the largest loss of the local maximum in the deposit currency and in per cents of the deposit;


I think, in effect, you're right about Max DD - it's the largest drawdown from the maximum attained during the test, shown as an amount in the deposit currency and also as a percentage of the maximum attained. I've never really sat and worked it out, just used it as a comparative figure.

Cheers,
Spin
SpindriftUK is offline   Reply With Quote
Old Feb 2, 2011, 4:55pm   #32
 
mark120169's Avatar
Joined Apr 2010
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by SpindriftUK View Post
Just finished the two tests and both faired slightly worse than the "b/e at 08:00" variants. Revised spreadsheets attached.
Interesting. I will stick with INT 30/30 AI RT and my +29 lockin for now. MMI was discovered later in 2010 after antonio's extensive testing. I discovered the AI in July 2010, was using it in August and started with it on here in September.

Also as MMI was decided on later in 2010 the backtest Data was available, so you could consider that as so much 2010 data was avaialble before live use is it "curve" fitted indicator wise? But I can't argue it is doing well in 2011 vs the MA40.

AI is quite a different method combining two indicators and has the edge so far in 2011 for live trading in RT mode vs MMI and MA40.

I will monitor the situation and consider again once Q1 is over.

M
mark120169 is offline   Reply With Quote
Old Feb 2, 2011, 5:47pm   #33
Joined Jun 2010
Re: Net Trap EA - variant backtesting

This is really interesting analysis, and does confirm that INT trumps NT every time.
From Goldfinger's data, I now trade 25INT30, 30INT40 and 45INT40, as I thought these gave increasing profit targets and fairly good returns over Goldfingers data set. Would it be possible to include these variants? Also, I am familiar with Marks AI, but haven't been following the MMI much. Would you mind adding a definitions section if it's not too much trouble.
Much appreciated.
DaxM is offline   Reply With Quote
Old Feb 2, 2011, 7:35pm   #34
Joined Jun 2010
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by SpindriftUK View Post
Just finished the two tests and both faired slightly worse than the "b/e at 08:00" variants. Revised spreadsheets attached.
Cheers Spin, great work.

Can you just confirm which SL and TP values you backtested, because in your spreadsheet you've got no INT variant with a 40 stop, and also you've got 22INT23 as 23TP and 22SL, when this particular variant in my INT spreadsheet is actually the opposite; 22TP and 23SL.

Just want to check exactly which variants you've selected, unless they're intentionally different from the "official" ones?
goldfinger777 is offline   Reply With Quote
Old Feb 2, 2011, 8:49pm   #35
Joined Dec 2006
Re: Net Trap EA - variant backtesting

SpindriftUK started this thread
Quote:
Originally Posted by goldfinger777 View Post
Cheers Spin, great work.

Can you just confirm which SL and TP values you backtested, because in your spreadsheet you've got no INT variant with a 40 stop, and also you've got 22INT23 as 23TP and 22SL, when this particular variant in my INT spreadsheet is actually the opposite; 22TP and 23SL.

Just want to check exactly which variants you've selected, unless they're intentionally different from the "official" ones?
I added the SL and TP columns to the spreadsheet in an attempt to avoid confusion as I was aware that I've gone against the standard in the past

So what's on the spreadsheet is what I've tested. I got the impression that quite a few people used 22sINT23l, which is why I included it. Also, the main purpose of the test was to compare MMI with AI, so I wasn't particularly analytical when choosing the variants.

Cheers,
Spin
SpindriftUK is offline   Reply With Quote
Old Feb 2, 2011, 9:29pm   #36
 
heatonfan's Avatar
Joined Mar 2009
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by SpindriftUK View Post
Just finished the two tests and both faired slightly worse than the "b/e at 08:00" variants. Revised spreadsheets attached.
Very nice work.

Probably time for me to buy a few pints for Antonio now.
heatonfan is offline   Reply With Quote
Old Feb 3, 2011, 6:42am   #37
Joined Oct 2010
Re: Net Trap EA - variant backtesting

Thanks Spin - interesting info...
mittons is offline   Reply With Quote
Old Feb 3, 2011, 7:44am   #38
Joined Dec 2006
Re: Net Trap EA - variant backtesting

SpindriftUK started this thread
Quote:
Originally Posted by mark120169 View Post
Interesting. I will stick with INT 30/30 AI RT and my +29 lockin for now. MMI was discovered later in 2010 after antonio's extensive testing. I discovered the AI in July 2010, was using it in August and started with it on here in September.

Also as MMI was decided on later in 2010 the backtest Data was available, so you could consider that as so much 2010 data was avaialble before live use is it "curve" fitted indicator wise? But I can't argue it is doing well in 2011 vs the MA40.

AI is quite a different method combining two indicators and has the edge so far in 2011 for live trading in RT mode vs MMI and MA40.

I will monitor the situation and consider again once Q1 is over.

M
I suppose any use of back test results to derive a strategy is going to feel a bit like curve fitting. Even forward tested strategies fall into this trap though, as the longer a particular thing works, the more it feels like it's going to continue to work forever. Today's forward test is tomorrow's back test!

I would really like to find a way to use the power of the EA to make a judgement as to what might work well on any given day. Sadly, I've yet to come with anything but if anybody has any good ideas in this area, please let me know! I would be happy to add some experimental code to the EA...

Cheers,
Spin
SpindriftUK is offline   Reply With Quote
Old Feb 3, 2011, 7:54am   #39
 
surferking's Avatar
Joined Oct 2010
Re: Net Trap EA - variant backtesting

Perhaps average trading range (ATR) checks could be added to calculate optimum size of TP and SL?

I have a could of ideas for adding indicator to the EA, will PM you...
surferking is offline   Reply With Quote
Old Feb 3, 2011, 7:56am   #40
 
surferking's Avatar
Joined Oct 2010
Re: Net Trap EA - variant backtesting

Spin,

I quick question if I may...

I recall I remember you added code to the EA to check for and delay trade until the service became available. Is this a MT4 service of which there will one instance per running application or is this a windows/dos service where there is only one available per instance of the operating system?

I'm trying to evaluate if running MT4 on multiple SLM accounts and an Oanda account will have an adverse effect if there's trade requests on each instance of MT4 at the same time.

I run Windows on Parallels so can run multiple instances no problem but its easier to do it all in one instance and just have multiple MT installations.

Cheers
surferking is offline   Reply With Quote
Old Feb 3, 2011, 8:41am   #41
Joined Dec 2006
Re: Net Trap EA - variant backtesting

SpindriftUK started this thread
Quote:
Originally Posted by surferking View Post
Spin,

I quick question if I may...

I recall I remember you added code to the EA to check for and delay trade until the service became available. Is this a MT4 service of which there will one instance per running application or is this a windows/dos service where there is only one available per instance of the operating system?

I'm trying to evaluate if running MT4 on multiple SLM accounts and an Oanda account will have an adverse effect if there's trade requests on each instance of MT4 at the same time.

I run Windows on Parallels so can run multiple instances no problem but its easier to do it all in one instance and just have multiple MT installations.

Cheers
Yes, it's an MT4 service, one instance per running application. You can certainly have more than one instance of MT4 running simultaneously in the same Windows instance. You just install each MT4 into a separate directory/folder and then have them all pointing at the same server/account combination. You have to be careful if you're running the same EA on different instances that you've got different trade_idents (magic numbers) for each instance of the EA, something which is true also when running the EA on more than one chart of the same pair in the same MT4 instance.

Cheers,
Spin
SpindriftUK is offline   Reply With Quote
Old Feb 3, 2011, 9:09am   #42
 
peterf1966's Avatar
Joined Feb 2010
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by SpindriftUK View Post
I suppose any use of back test results to derive a strategy is going to feel a bit like curve fitting. Even forward tested strategies fall into this trap though, as the longer a particular thing works, the more it feels like it's going to continue to work forever. Today's forward test is tomorrow's back test!

I would really like to find a way to use the power of the EA to make a judgement as to what might work well on any given day. Sadly, I've yet to come with anything but if anybody has any good ideas in this area, please let me know! I would be happy to add some experimental code to the EA...

Cheers,
Spin
Greg,

I pulled a long trade the EA opened on19/01/11 Candles were forming with spike tops (inverted hammers)- 4 in a row I think, followed by a couple of doji candles. Looked iffy so I pulled it, and it saved a loss. The EA could check high and low of a candle against the close price to detect this...but it's a lot of +D, and I wonder if it's making things too complex - remember KISS!
peterf1966 is offline   Reply With Quote
Old Feb 3, 2011, 10:05am   #43
 
mark120169's Avatar
Joined Apr 2010
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by peterf1966 View Post
Greg,

I pulled a long trade the EA opened on19/01/11 Candles were forming with spike tops (inverted hammers)- 4 in a row I think, followed by a couple of doji candles. Looked iffy so I pulled it, and it saved a loss. The EA could check high and low of a candle against the close price to detect this...but it's a lot of +D, and I wonder if it's making things too complex - remember KISS!
I agree the EA could get too complicated and spend more time finding reasons not to trade.

Don't forget with a new target of 125 pips a month, which is 1500 p/a. On a 30 stop @ 2% risk per trade this is a 100% gain p/a WITHOUT weekly compounding.

We are on track aleady to acheive this via two methods, both quite simple. Apart for AI rules I switch everything else off. I love the AI, the trend is your friend and it works and the track record is holding up for 13 months of trading data now.

Losses will happen, but AI RT drawdown vs compounding has an excellent ratio so far too.

KISS is the way to go.

M
mark120169 is offline   Reply With Quote
Old Feb 4, 2011, 7:54pm   #44
Joined Jun 2010
Re: Net Trap EA - variant backtesting

Quote:
Originally Posted by mark120169 View Post
I agree the EA could get too complicated and spend more time finding reasons not to trade.

Don't forget with a new target of 125 pips a month, which is 1500 p/a. On a 30 stop @ 2% risk per trade this is a 100% gain p/a WITHOUT weekly compounding.

We are on track aleady to acheive this via two methods, both quite simple. Apart for AI rules I switch everything else off. I love the AI, the trend is your friend and it works and the track record is holding up for 13 months of trading data now.

Losses will happen, but AI RT drawdown vs compounding has an excellent ratio so far too.

KISS is the way to go.

M
Yep, using an EA definitely helps to keep things simple, especially with regard to set and forget. I'm a huge fan of EAs!
goldfinger777 is offline   Reply With Quote
Old Feb 10, 2011, 10:25pm   #45
Joined Jan 2011
Re: Net Trap EA - variant backtesting

Hi guys i'm very new to using an EA but so far ive had two trades in 2 days and i like what i see so far. i'm trying to work out how to make the EA trail the stops for example or to make it take profit at 30 pips , but not 100% how to do it. I changed the TP variable on the input from 50 to 30 but it didnt seem to work.
Any advice would be grealty appreciated.
Thankyou
Almontana73 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search

Similar Threads
Thread Thread Starter Forum Replies Last Post
Net Trap Variant live results from 17/05/10 mark120169 Forex Net Trap 572 Dec 5, 2011 10:54pm
Net Trap Tools mark120169 Forex Net Trap 10 Dec 11, 2010 9:02pm
Forex Net Trap SA Trader Forex Net Trap 6 Jun 28, 2010 1:44pm
Net Trap Johnnysportster First Steps 5 Mar 30, 2010 9:08am

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)